OpenAI's GPT Store Launches Next Week
adsby.co
adsby.co
Their suggested replacement is a gpt-3.5-turbo-instruct model and those pale in comparison to text-davinci-003 original. It's a shame. Some of my services will stop working (in the sense that users won't care to use the service). The local LLM replacements like Mixtral-8x7B are not as good as text-davinci-003 but they're definitely better than a gpt-3.5-turbo.
I am curious to what kind of task load is benefitting greatly from text-davinci-003?
They could just say those models are less cost/efficiency optimized and raise the costs to cover the expenses + even an extra margin.
Also, All custom gpt seem to constantly forget their instructions . They would respond correctly one or two times in the playground and revert back to previous setting.
There is already a post about the email from OpenAI. I am flagging the post.